The Electron-Ion Collider (EIC) is a next-generation facility designed to probe the fundamental structure of matter: nucleons and nuclei. I provide a brief overview of the physics opportunities at the EIC.

A groundbreaking advancement in nuclear physics is now being realized: the use of electron scattering to investigate the structure of short-lived unstable nuclei, previously deemed impossible. Electron scattering, which relies on the well-understood electromagnetic interaction, is a powerful and unambiguous method for probing nuclear structure.

Despite the strong desire to apply electron scattering to unstable nuclei, it had long been considered unfeasible due to the difficulty in preparing sufficient quantities of short-lived unstable nuclei as targets. These nuclei decay quickly and cannot be accumulated into thick targets, as required in conventional electron scattering experiments.

To overcome this challenge, our team developed the Self-Confining RI Ion Target (SCRIT) system and constructed a dedicated electron scattering facility at RIKEN. SCRIT allows for the creation of a stationary target with unstable ions trapped by the electric force from the electron beam and an axial potential well generated by surrounding electrodes. This enables a high luminosity even with an extremely low number of target ions, about ten orders of magnitude fewer than in conventional setups.

After nearly two decades of development, we successfully conducted the world’s first electron scattering experiment using an online-produced unstable nucleus. We chose ¹³⁷Cs as the target nucleus due to its favorable extraction and transport efficiency.

The observed elastic scattering angular distribution from ¹³⁷Cs matched the expected curve derived from its nuclear size, confirming the validity of the method. Although ¹³⁷Cs has a relatively long half-life (～30 years), the entire experiment effectively mimics conditions for truly short-lived nuclei.

Following this success, facility upgrades are underway to increase the electron beam power by a factor of 100, aiming to measure charge distributions of various short-lived unstable nuclei. A key future goal is to measure the charge distribution of ¹³²Sn, a doubly magic nucleus in the unstable region.

We discuss a machine learning application to the four-dimensional string vacua, so-called the string landscape. The string landscape can be characterized by the topological data in string theory, but it is difficult to find out the realistic Standard Model of particle physics from vast number of string vacua. As a concrete application of machine learning to the string landscape, we deal with the deep autoencoder to type IIA intersecting D-brane models.

The speed of sound serves as a key probe of matter under extreme conditions, such as the early universe and the interiors of neutron stars. Until now, relativistic quantum field theories, such as QCD, suggested an upper limit of c/√3, known as the conformal bound. Recently, first-principles lattice simulations of two-color QCD, where the sign problem is absent, shows that the speed of sound can exceed this bound in high-density regimes. This provides the first lattice evidence of conformal bound violation, consistent with recent neutron star observations. Future advances in both computational methods and astrophysical measurements are expected to further clarify the ultimate limit of sound velocity in nature.